Questo contenuto non è disponibile nella lingua selezionata.
Chapter 10. Detecting duplicate messages
You can configure the broker to automatically detect and filter duplicate messages. This means that you do not have to implement your own duplicate detection logic.
Without duplicate detection, in the event of an unexpected connection failure, a client cannot determine whether a message it sent to the broker was received. In this situation, the client might assume that the broker did not receive the message, and resend it. This results in a duplicate message.
For example, suppose that a client sends a message to the broker. If the broker or connection fails before the message is received and processed by the broker, the message never arrives at its address. The client does not receive a response from the broker due to the failure. If the broker or connection fails after the message is received and processed by the broker, the message is routed correctly, but the client still does not receive a response.
In addition, using a transaction to determine success does not necessarily help in these cases. If the broker or connection fails while the transaction commit is being processed, the client is still unable to determine whether it successfully sent the message.
In these situations, to correct the assumed failure, the client resends the most recent message. The result might be a duplicate message that negatively impacts your system. For example, if you are using the broker in an order-fulfilment system, a duplicate message might mean that a purchase order is processed twice.
The following procedures show how to configure duplicate message detection to protect against these types of situations.
10.1. Configuring the duplicate ID cache
To enable the broker to detect duplicate messages, producers must provide unique values for the message property _AMQ_DUPL_ID
when sending each message. The broker maintains caches of received values of the _AMQ_DUPL_ID
property. When a broker receives a new message on an address, it checks the cache for that address to ensure that it has not previously processed a message with the same value for this property.
Each address has its own cache. Each cache is circular and fixed in size. This means that new entries replace the oldest ones as cache space demands.
The following procedure shows how to globally configure the ID cache used by each address on the broker.
Procedure
-
Open the
<broker_instance_dir>/etc/broker.xml
configuration file. Within the
core
element, add theid-cache-size
andpersist-id-cache
properties and specify values. For example:<configuration> <core> ... <id-cache-size>5000</id-cache-size> <persist-id-cache>false</persist-id-cache> </core> </configuration>
id-cache-size
Maximum size of the ID cache, specified as the number of individual entries in the cache. The default value is 20,000 entries. In this example, the cache size is set to 5,000 entries.
NoteWhen the maximum size of the cache is reached, it is possible for the broker to start processing duplicate messages. For example, suppose that you set the size of the cache to
3000
. If a previous message arrived more than 3,000 messages before the arrival of a new message with the same value of_AMQ_DUPL_ID
, the broker cannot detect the duplicate. This results in both messages being processed by the broker.persist-id-cache
-
When the value of this property is set to
true
, the broker persists IDs to disk as they are received. The default value istrue
. In the example above, you disable persistence by setting the value tofalse
.
Additional resources
- To learn how to set the duplicate ID message property using the AMQ Core Protocol JMS client, see Using duplicate message detection in the AMQ Core Protocol JMS client documentation.
10.2. Configuring duplicate detection for cluster connections
You can configure cluster connections to insert a duplicate ID header for each message that moves across the cluster.
Prerequisites
- You should have already configured a broker cluster. For more information, see Section 14.2, “Creating a broker cluster”
Procedure
-
Open the
<broker_instance_dir>/etc/broker.xml
configuration file. Within the
core
element, for a given cluster connection, add theuse-duplicate-detection
property and specify a value. For example:<configuration> <core> ... <cluster-connections> <cluster-connection name="my-cluster"> <use-duplicate-detection>true</use-duplicate-detection> ... </cluster-connection> ... </cluster-connections> </core> </configuration>
use-duplicate-detection
-
When the value of this property is set to
true
, the cluster connection inserts a duplicate ID header for each message that it handles.